Discovery Labs
Learn how we unravel the hidden complexities of logistics and apply state-of-the-art in Operations Research to real-world problems. At VersaFleet, we implement Deep Technology that thrive in the chaos of everyday logistics.

Our Unsung Everyday Heroes
Crafting “the perfect route plan” while considering multiple constraints to successfully fulfill orders is not a simple feat, especially on a daily basis. Planning work typically takes 2-4 hours’ each time. Mistakes are incredibly costly because they result in multiple trips, wasted movement and ultimately: frustrated customers.
With logistics & supply chain management at the very core of global trade, there are ever increasing considerations that must be incorporated into successful route plans. Rising customer expectations used to ‘on demand’ services has only increased the pressure and expectations demanded of professional logisticians.
Despite these challenges, working tirelessly every day to come up with route schedules that work, are ground operations staff, route planners, schedulers, logistics & transport managers. Most of them work through these complex logical problems “by hand”, through sheer brain power and relying solely on their years of experience!
While we take a moment to applaud their efforts, we must acknowledge that there is more science than art in this form of knowledge work. This planning work can be easily solved by AI in seconds, instead of hours if by humans.

Unraveling the Complexity
We realized that in the eyes of a professional route planner, no two cities have the same road networks or operate in exactly the same way . Different cities have their own geographical contexts, unique regulatory considerations, and just “special constraints” that must be respected when optimising route plans. These are over and above each businesses’ very practical last mile requirements like time-windows and location opening hours.
Added to that, although it is common to outsource transportation to 3rd-party Logistics Service Providers, cost models for transporters differ widely, even among 3PLs to the same customer! There is typically no universally recognised standard. This means that an intent to “reduce transport costs” will have vastly differing meanings to businesses in different cities, and these cannot always be generalised.
A route optimisation engine specifically designed for the last mile must be sufficiently flexible and configurable to accommodate and adapt to this wide spectrum of constraints and cost models.
As if VRP on its own was not a complex enough logic problem, professional route planners are expected to factor in multiple cost-models and inter-city variables as well – all solved by hand no less!
Progressing beyond Existing Methods
Computing the Vehicle Routing Problem (VRP) for the Everyman.
VRP has been widely recognised as a NP-Hard problem. This means that as a complexity class, VRP is non-deterministic and in polynomial time. In other words, it is at least as hard as the hardest problems in computer science. In fact, VRP is a prime candidate to solve by quantum computing.
While there have been decades of scholarship in this field, it is not widely accessible in a form that can be understood by actual practitioners.
Linear programming (LP)
An LP approach, for example as mixed integer formulation, aims to solve for a “perfectly optimal” solution each and every time!
However, there is a limit as to how large a problem-set can be solved in this manner. Anything more than 100 delivery points is generally assumed to be too large for LP to solve on classical computing.
In practice, attempting to solve VRP by LP on a daily basis is not feasible because it will take hours and days to run each time, even on a high-performance grid computing network.
The pseudo-code alone, while well published, is practically too complex for the everyman in supply chain to use on a daily basis.
Partition, Map, Reduce (Zoning)
The “go to” method used by most planners when crafting route schedules, popularly known as ‘Zoning’ or ‘planning by territory’.
This is easy to implement, but only feasible only when businesses have density on the scale of national post carriers.
Without high distribution densities, this approach often results in over-utilization, under-utilisation or even un-utilisation of resources, which is less than optimal.
Intuitively, a company with 4 vehicles cannot simply assign “one vehicle per zone”, because distribution densities will likely shift day to day, resulting in under- or over-capacity.
Genetic algorithms (meta-heuristics)
The cutting-edge and state-of-the-art, heuristics based approaches have been proven to reliably produce ‘near-optimal’ results that satisfice real-world requirements.
In practice though, it is very challenging to implement even the best algorithms without factoring how the end-user will handle the multiple and ever-changing constraints that must be taken into consideration for every route plan, every day.
The critical challenge is to build an intuitive interface for a route-planner to input real-world constraints easily, without necessarily having a PhD in Operations Research!
This is where VersaFleet continues to invest heavily into our product development, not just for a powerful optimisation engine, but for an intuitive overall User Experience.

Collaborating closely with AI Enabling Technology
Incorporate the latest in AI technology into VersaFleet’s route solver, adding more depth and dimension to better serve the needs of businesses and their geographical context.
Constantly enhancing the TMS technology to deliver a smooth experience to our unsung heroes so that they can execute their jobs faster, leaner and more efficiently.
Technobabble made easy
Keep up to date with the latest technological news in the logistics world.

[TMS for Non-Logistics Industries] Food Manufacturing Part 1: Why should TMS be Considered?
While a Transport Management System (TMS) is greatly associated with the logistics industry, the logistics industry is not the only one that is in need

Big Data & Logistics (Part III): Roadblocks and Considerations
After covering the what and the why of Big Data in our first post, we moved on to the how in our second post titled

Big Data & Logistics (Part II): How to implement?
In our previous post titled “Big Data & Logistics (Part I): Breaking Down Big Data“, we covered the what and the why of Big Data.

Navigating The Vehicle Routing Problem
Most transporters constantly try to improve their efficiency and productivity as they grow in order to maintain or reduce their costs and expenses. In order

Big Data & Logistics (Part I): Breaking Down Big Data
The term ‘Big Data’ has always served to confuse with its broad ambit and unclear usage. At its core, Big Data is the utilisation of

FMS or TMS: Choosing The Logistics Operations Software You Need
Transporters are starting to recognise the need for technology to improve efficiency and productivity in the face of the rising popularity of e-commerce and consumers’